Skip to main content
. 2017 Jul 31;45(17):e156. doi: 10.1093/nar/gkx681

Table 1. The five different types of NN used in the paper.

No. Model Layer1#node Layer2#node #Parameters (million)
1 Dense 796 X 7.52 M
2 Dense 100 X 0.95 M
3 Dense 796 100 7.59 M
4 PPI/TF+dense 696 + 100 dense X 1.01 M
5 PPI/TF+dense 696 + 100 dense 100 1.08 M

Note that No.1 and No.2 represent the same architecture (Figure 1A) using different number of hidden layer nodes. The 696 in the PPI/TF models is from 348 TFs + 348 PPI groups. The number of nodes in the Dense models corresponds to the number of nodes in PPI/TF models for comparison. The additional 100 dense in each model are manually selected.